ianmcook commented on a change in pull request #9745:
URL: https://github.com/apache/arrow/pull/9745#discussion_r601596482
##########
File path: r/R/dplyr.R
##########
@@ -689,17 +741,77 @@ abandon_ship <- function(call, .data, msg = NULL) {
eval.parent(call, 2)
}
-arrange.arrow_dplyr_query <- function(.data, ...) {
+arrange.arrow_dplyr_query <- function(.data, ..., .by_group = FALSE) {
+ call <- match.call()
+ exprs <- quos(...)
+ if (.by_group) {
+ # when the data is is grouped and .by_group is TRUE, order the result by
+ # the grouping columns first
+ exprs <- c(quos(!!!dplyr::groups(.data)), exprs)
+ }
+ if (length(exprs) == 0) {
+ # Nothing to do
+ return(.data)
+ }
.data <- arrow_dplyr_query(.data)
- if (query_on_dataset(.data)) {
- not_implemented_for_dataset("arrange()")
+ # find and remove any dplyr::desc() and tidy-eval
+ # the arrange expressions inside an Arrow data_mask
+ sorts <- vector("list", length(exprs))
+ descs <- logical(0)
+ mask <- arrow_mask(.data)
+ for (i in seq_along(exprs)) {
+ x <- find_and_remove_desc(exprs[[i]])
+ exprs[[i]] <- x[["quos"]]
+ sorts[[i]] <- arrow_eval(exprs[[i]], mask)
+ if (inherits(sorts[[i]], "try-error")) {
+ msg <- paste('Expression', as_label(exprs[[i]]), 'not supported in
Arrow')
+ return(abandon_ship(call, .data, msg))
+ }
+ names(sorts)[i] <- as_label(exprs[[i]])
+ descs[i] <- x[["desc"]]
}
- # TODO(ARROW-11703) move this to Arrow
- call <- match.call()
- abandon_ship(call, .data)
+ .data$arrange_vars <- c(sorts, .data$arrange_vars)
+ .data$arrange_desc <- c(descs, .data$arrange_desc)
+ .data
}
arrange.Dataset <- arrange.ArrowTabular <- arrange.arrow_dplyr_query
+# Helper to handle desc() in arrange()
+# * Takes a quosure as input
+# * Returns a list with two elements:
+# 1. The quosure with any wrapping parentheses and desc() removed
+# 2. A logical value indicating whether desc() was found
+# * Performs some other validation
+find_and_remove_desc <- function(quosure) {
+ expr <- quo_get_expr(quosure)
+ descending <- FALSE
+ if (length(all.vars(expr)) < 1L) {
+ stop(
+ "Expression in arrange() does not contain any field names: ",
+ deparse(expr),
+ call. = FALSE
+ )
+ }
+ while (identical(typeof(expr), "language") && is.call(expr)) {
Review comment:
On bad inputs, most of the errors will be thrown by
`rlang::eval_tidy()`. For example, if you pass nonexistent column names, it
gives errors like this:
```
> iris %>% Table$create() %>% arrange(z)
Error in eval_tidy(expr, mask) : object 'z' not found
```
but there are some weird cases like this
```
> x <- iris %>% Table$create() %>% arrange(q)
# no error (q is a function)
> x
Error: object of type 'closure' is not subsettable
```
I'll add some more checks and test code for bad inputs like that.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
